How to smooth edges of polygons?

I created a image (uint8) with multiple filled polygons. First, I used the vision.ShapeInserter with 'Antialiasing' but this was not enough. Also insertShape + 'SmoothEdges' produced insufficient smooth edges. Are there any other ways how I can fix that?
Looking forward to your suggestions. Thanks!

답변 (2개)

Andrew Chen
Andrew Chen 2017년 10월 12일

0 개 추천

Perhaps imopen with a disk structuring element will provide the functionality you are looking for?
Ex:
se = strel('disk',5);
new_img = imopen(original_image,se);
Hope this helps!

댓글 수: 3

m975
m975 2017년 10월 13일
Hi Andrew and thanks for your answer! As you can see it did not really help. I guess the edges are too fine. Are there any other suggestions?
Andrew Chen
Andrew Chen 2017년 10월 13일
Did Image Analyst's answer get you what you wanted? Otherwise, maybe you could clarify what you mean by "smooth edges". What in your mind constitutes a sufficiently smooth edge?
m975
m975 2017년 10월 13일
For example, in this image polygons show proper edges without these awful "stairs" when the shapes get rotated. By the way, this image was drawn in photoshop. But now I have to create a lot of stimuli in MATLAB by using coordinates for my polygons. Consequently, I am looking for a MATLAB routine to generate images of same quality. I hope this explanation helps. Thanks!

댓글을 달려면 로그인하십시오.

Image Analyst
Image Analyst 2017년 10월 13일

0 개 추천

You can smooth the edges with a filter. See attached.

댓글 수: 8

m975
m975 2017년 10월 13일
Thanks for the code! But is there is also an easier and more fundamental solution? Due to the fact that I create these polygons in MATLAB directly I cannot understand why the output of such simple shapes can be so bad. Could maybe a vector based format be the right way? Unfortunately, I did not find such a format for imwrite.
Image Analyst
Image Analyst 2017년 10월 13일
I don't know how they were created and what they are. If they are images, are they grey scale, or binary? If they're binary images, you can smooth them nicely by blurring with conv2() or imfilter() and then thresholding.
m975's "Answer" moved here because it's not an "Answer" to the original question, but a reply to me.
I opened an image with
image = zeros(PicHeight,PicWidth,3,'uint8') + 255;
and inserted polygons with the vision.ShapeInserter. Finally, I saved the images with imwrite as PNG.
OK, then blurring should work, like I said. Did you try it? If not, why not?
blurredImage = imfilter(grayImage, ones(9)/81);
m975
m975 2017년 10월 13일
Because I thought you need more information about the image. Now I tried it but it looks too blurish.
Image Analyst
Image Analyst 2017년 10월 13일
It looks binary, so just threshold it. The corners should then be more rounded off.
I processed the image as follows but it did not work.
level = graythresh(I);
BW = im2bw(I, level);
You can still see these toothed edges:
m975
m975 2017년 10월 19일
Finally, the easiest workaround was to create very big images (5000 x 5000 px) and, subsequently, resize them to the favored stimulus size. Now edges look sexy again!

댓글을 달려면 로그인하십시오.

카테고리

도움말 센터File Exchange에서 Get Started with Computer Vision Toolbox에 대해 자세히 알아보기

질문:

2017년 10월 12일

댓글:

2017년 10월 19일

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by